Weak symmetry in naturally reductive homogeneous nilmanifolds (Q1880978)

\textit{A. Selberg} [J. Indian Math. Soc., n. Ser. 20, 47-87 (1956; Zbl 0072.08201)] introduced the notion of a \textit{weakly symmetric} Riemannian manifold. He showed that such a manifold is ``commutative'', i.e., that the algebra of all invariant differential operators which are invariant under the connected component of the isometry group is a commutative algebra. He also asked whether a manifold which is commutative in this sense must necessarily be a weakly symmetric space. \textit{D. N. Akhiezer} and \textit{E. B. Vinberg} [Transform. Groups 4, No.1, 3-24 (1999; Zbl 0916.53024)] proved that this is true provided the manifold is homogeneous under a reductive group. On the other hand, \textit{J. Lauret} [Bull. Lond. Math. Soc. 30, No.1, 29-36 (1998; Zbl 0921.22007)] proved that there do exist nilmanifolds which are commutative in the above sense, but not weakly symmetric. In this article it is verified, using a case-by-case check, that a nilmanifold which is a commutative space \textit{and} ``naturally reductive'', must be a weakly symmetric space. Thus it is not by coincidence that the aforementioned examples of Lauret fail to be naturally reductive.